Keratoconjunctivitis sicca in dogs under primary veterinary care in the UK: an epidemiological study

O bjectives : To estimate the frequency and breed-related risk factors for keratoconjunctivitis sicca (KCS) in dogs under UK primary veterinary care. M ethOds : Analysis of cohort electronic patient record data through the VetCompass Programme. Risk factor analysis used multivariable logistic regression. R esults : There were 1456 KCS cases overall from 363,898 dogs [prevalence 0.40%, 95% confidence interval (CI) 0.38–0.42] and 430 incident cases during 2013 (1-year incidence risk 0.12%, 95% CI 0.11–0.13). Compared with crossbreds, breeds with the highest odds ratio (aOR) for KCS included American cocker spaniel (aOR 52.33: 95% CI 30.65–89.37), English bulldog (aOR 37.95: 95% CI 26.54–54.28), pug (aOR 22.09: 95% CI 15.15–32.2) and Lhasa apso (aOR 21.58: 95% CI 16.29– 28.57). Conversely, Labrador retrievers (aOR 0.23: 95% CI 0.1–0.52) and border collie (aOR 0.30: 95% CI 0.11–0.82) had reduced odds. Brachycephalic dogs had 3.63 (95% CI 3.24–4.07) times odds compared to mesocephalics. Spaniels had 3.03 (95% CI 2.69–3.40) times odds compared to non-spaniels. Dogs weighing at or above the mean bodyweight for breed/sex had 1.25 (95% CI 1.12–1.39) times odds compared to body weights below. Advancing age was strongly associated with increased odds. of eye in breeding strategies could aim of conformation.

INTRODUCTION schools that used the Veterinary Medical Data Program to report predisposition to KCS in several dog breeds spanning a 12-year period (1964)(1965)(1966)(1967)(1968)(1969)(1970)(1971)(1972)(1973). Kaswan and Salisbury (1990) also used the Veterinary Medical Data Program to propose several breed predispositions with data that spanned a 34-year period . The results of these two studies concurred on six breed predispositions: English bulldog, West Highland white terrier, pug, cocker spaniel, pekingese and miniature schnauzer. Other breeds reported as predisposed by one or other study included cavalier king Charles spaniel, Lhasa apso, shih-tzu, blood hound, Boston terrier, Samoyed, Yorkshire terrier and English springer spaniel , Helper 1996. A 1985 study in the UK with 200 referral cases of primary and secondary KCS collected over a 9-year period reported that just over onethird of the cases were West Highland white terriers, suggesting a predisposition in this breed (Sansom & Barnett 1985). A later 2007 study, also from the UK, included 229 referral primary KCS cases spanning an 8-year period and suggested predispositions in English cocker spaniel, cavalier king Charles spaniel, West Highland white terrier and shih-tzu breeds (Sanchez et al. 2007). However, the prevalence and incidence risk of KCS in the general population of dogs around the world have remained largely unreported because the aforementioned studies focused on referral rather than primary care practice populations (Bartlett et al. 2010). In addition, these previous studies either contained relatively few cases or required a span of several years to capture larger case numbers (Sanchez et al. 2007, Sansom & Barnett 1985, sometimes extending to several decades (Helper 1996, Kaswan et al. 1991. Ideal study populations for epidemiological studies that aim to report on disease risk should be large and representative enough to comfortably allow for detection of adequate cases, show variation in the putative risk associated factors and span a short and defined time frame (e.g. 1 year) (Dohoo et al. 2009). In addition, there is a growing call for studies that include only primary care cases to avoid a referral bias and to promote generalisability to the wider dog population (Bateson 2010, McGreevy & Nicholas 1999, O'Neill et al. 2014a. Unfortunately, until recently, lack of availability of access to clinical data on large enough populations of primary care caseloads has prevented estimation of prevalence and incidence risk for many important disorders such KCS, and consequently many breed-related associations remained unreported. However, in recent times, specialised computer programmes have enabled data sharing into national epidemiological projects from a large number of participating primary-care veterinary practices within a geographical area (PETscan 2020, SAVS-NET 2020, VetCompass Australia 2020, VetCompass 2021). In the UK, the VetCompass™ programme collates de-identified electronic patient record (EPR) data on over 15 million companion animals from over 1800 primary-care veterinary practices for epidemiological research (VetCompass 2021).
This study aimed to estimate the frequency of KCS in dogs in the UK using data extracted from a large number of animals under primary veterinary care within 2013 using the VetCompass Programme. The study additionally aimed to evaluate breed and breedrelated elements, such as brachycephalia, as risk factors for KCS.

MATERIALS AND METHODS
The cohort of dogs attending VetCompass practices was used to estimate the prevalence, incidence risk and risk factors for KCS during 2013 (Pearce 2012). The sampling frame for the current study included dogs under veterinary care within the VetCompass database from January 1, 2013 to December 31, 2013. Dogs "under veterinary care" were defined as having at least one EPR recorded from January 1, 2013 to December 31, 2013 and/or any dog with at least one EPR both before and after 2013. Sample size calculations estimated that 8625 dogs of a specific breed or group (e.g. brachycephalic or spaniel) and 77,619 dogs that were not of this type would be required to detect an odds ratio (OR) of 1.5 times or greater for KCS assuming a 0.5% prevalence of KCS in the non-type dogs [9:1 ratio of non-type: type, two-sided 95% confidence interval (CI), 80% power] (Epi Info 7 CDC 2019). Ethical approval was granted by the Royal Veterinary College Ethics and Welfare Committee (reference number 2016/U91). Demographic information on each dog was extracted from Vet-Compass on breed, date of birth, sex, neuter status and bodyweight (O'Neill et al. 2014b).
The inclusion criteria for KCS required either a recorded diagnosis of KCS or synonym (e.g. dry eye) and/or at least one Schirmer's tear test-1 <15 mm/minute (or stated to be subnormal) with a concomitant recommendation specifically for topical cyclosporine therapy. Exclusion criteria included dogs that initially met the inclusion criteria but where KCS diagnosis was subsequently recorded as ruled out. For the current study, a KCS case was defined as any dog that met the KCS inclusion criteria described above for a condition that was present during 2013. Case finding involved initial screening of all EPRs of all dogs to identify candidate KCS cases by searching the clinical free text notes field for: "dry eye", kcs*, keratoconj*, keratocon-junctivitis~2, sicca~1, optimmune~1 and the treatment field for: cyclosp*, tacro*, optim*. The EPRs of all candidate cases were manually reviewed in detail to verify case inclusion and the date of first diagnosis was extracted. Confirmed KCS cases were categorised as incident (newly diagnosed during 2013) or preexisting (first diagnosed before 2013).
Risk factor analysis grouped all dogs with confirmed KCS as KCS cases and all remaining dogs as non-cases. A breed variable included all individual breeds with ≥10 KCS cases or ≥1500 dogs in the overall study to allow focus on commonly affected breed types and to facilitate statistical power for the individual breed analyses (Scott et al. 2012). A purebred variable categorised currently recognised breeds as purebreds and all remaining types as crossbreds (Dog Breed Info 2019, Irion et al. 2003, The Kennel Club 2020b. Purebreds were further categorised by UK Kennel Club breed-recognition (recognised/not recognised) and UK Kennel Club breed group (gundog, hound, pastoral, terrier, toy, utility, working) (The Kennel Club 2020b). Purebreds were also separately categorised based on skull conformation (dolichocephalic, mesocephalic and brachycephalic), spaniel status, poodle status and dachshund status (Appendix S1). Neuter status was defined at the final available EPR. Age was defined as the age (years) at December 31, 2013 and was British Small Animal Veterinary Association categorised into five groups: < 3.0, 3.0 to <6.0, 6.0 to <9.0, 9.0 to <12.0 and ≥12.0. Adult bodyweight was defined as the mean of all bodyweight (kg) values recorded for each dog after reaching 18 months and was categorised into five groups <10.0, 10.0 to <20.0, 20.0 to <30.0, 30.0 to <40.0 and ≥40.0. The mean adult bodyweight was calculated for all breeds where information available for at least 100 dogs and each dog was categorised as lower than or "at or above" their relevant breed/ sex mean bodyweight. This variable allowed the effect of adult bodyweight to be assessed within each breed/sex combination (O'Neill et al. 2018).
Following data checking and cleaning in Excel (Microsoft Office Excel 2013, Microsoft Corp.), analyses were conducted using Stata Version 13 (Stata Corporation). One-year period prevalence with 95% CIs described the probability of being a KCS case at any time during the 1-year 2013 study period.
One-year incidence risk with 95% CIs described the probability of being newly diagnosed with KCS during 2013. The CI estimates were derived from standard errors, based on approximation to the binomial distribution (Kirkwood & Sterne 2003). Descriptive statistics characterised demography separately for the case and non-case dogs. Binary logistic regression modelling was used to evaluate univariable associations between risk factors [purebred,breed,skull conformation,dachshund type,spaniel type,poodle type,Kennel Club Breed Group,adult (>18 months) bodyweight (kg), bodyweight relative to breed/sex mean, age, sex and neuter] and KCS diagnosis. Because breed was a factor of primary interest for the study, purebred status, breed, skull conformation, dachshund type, spaniel type, poodle type, Kennel Club Breed Group (variables that are highly collinear with breed) and adult bodyweight (a defining characteristic of individual breeds) were excluded from the initial breed-based multivariable modelling. Instead, each of these variables individually replaced the breed variable in the final breed-based model to evaluate their effects after taking account of the other confounding variables. Risk factors with liberal associations in univariable modelling (P < 0.2) were taken forward for multivariable evaluation. Model development used manual backwards stepwise elimination. Pair-wise interaction effects were evaluated for the final model variables and confounding effects from dropped variables were assessed by individual re-introduction to the final model. Clinic attended was evaluated as a random effect in the final model (Dohoo et al. 2009). The area under the receiver operating characteristic (ROC) curve and the Hosmer-Lemeshow test were used to evaluate the quality of the model fit and discrimination (non-random effect model) (Dohoo et al. 2009, Hosmer et al. 2013. Statistical significance was set at P < 0.05. Univariable ORs are reported as OR whereas multivariable ORs are reported as adjusted OR (aOR).
Univariable logistic regression modelling identified 10 variables that were liberally associated with KCS and were further evaluated using multivariable logistic regression modelling: purebred, breed, skull conformation, spaniel type, poodle type, Kennel Club Breed Group, adult (>18 months) bodyweight (kg), bodyweight relative to breed/sex mean, age and neuter (Tables 1 and 2). The final breed-based multivariable model retained three risk factors: breed, age and bodyweight relative to breed/sex mean. No biologically significant interactions were identified in the final model. The final unclustered model showed acceptable model fit (Hosmer-Lemeshow test statistic: P = 0.102) and discrimination (area under the ROC curve: 0.880). The mixed effects model with clinic entered as a random effect was a better model of the data than the fixed effects model (P < 0.001) and these mixed effects results are reported. The intraclass correlation coefficient (rho) indicated that 3.03% of the variation in the data was due to clustering at the veterinary clinic level. After accounting for the effects of the other variables evaluated, 22 breeds showed increased adjusted  (Table 3).
After replacing breed from the final breed-based mixed-effects multivariable model, purebred dogs had 5.13 (95% CI 4.15-6.35, P < 0.001) times the adjusted odds of KCS compared with crossbreds. Brachycephalic types had 3.63 (95% CI 3.24-4.07, P < 0.001) times the adjusted odds compared with mesocephalic types. Spaniel types had 3.03 (95% CI 2.69-3.40, P < 0.001) times the adjusted odds compared with dogs that were not spaniel type. Compared with breeds that were not recognised by the Kennel Club, all Kennel Club breed groups except for the pastoral group had higher adjusted odds of KCS. Dogs in lower bodyweight groups had higher adjusted odds of KCS: dogs weighing 10.0 to <20.0 kg had 5.49 (95% CI 4.26-7.08, P < 0.001) times the adjusted odds compared with dogs weighing 30.0 to <40.0 kg (Table 4). Dachshund type and poodle type were not associated with the adjusted odds of KCS following multivariable modelling.

DISCUSSION
A large and representative sample of affected and unaffected animals are needed to report accurately on the prevalence, incidence and risk factors of a disorder, and, ideally, these data should span a period that is as short as possible (Dohoo et al. 2009). Achieving such a volume of data to report on relatively uncommon disorders is an arduous task and explains why previous studies of KCS that included over 200 animals required periods from 8 up to 34 years to accrue adequate case numbers (Helper 1996, Kaswan , Kaswan et al. 1991, Sanchez et al. 2007, Sansom & Barnett 1985. Moreover, application of referral populations will include an inevitable referral bias that limits generalisability to the wider dog population (Bartlett et al. 2010). Although each of the aforementioned studies on KCS has been instrumental in building our understanding of The breeds with the highest predispositions (i.e. aORs) in the present primary-care study were American cocker spaniel, English bulldog, pug, Lhasa apso, cavalier king Charles spaniel, English bull terrier, English cocker spaniel, basset hound, West Highland white terrier, shih-tzu and king Charles spaniel. The breed results from the present study and the studies by Sansom and Barnet (1985), Kaswan and Salisbury (1990), Helper (1996) and Sanchez et al. (2007) coincide substantially to offer strong evidence for predisposition to KCS in at least six breeds: English cocker spaniel, American cocker spaniel, pug, English bulldog, cavalier king Charles spaniel and West Highland white terrier.
Awareness of breed predisposition can be linked to positive actions to mitigate these disorders in commonly affected breeds. For example, a study demonstrated that frequent eye examinations coupled with breeding control strategies were associated with decreased incidence of hereditary cataract and progressive retinal atrophy in dachshunds in Germany (Koll et al. 2017). Yet, despite the accumulating evidence of breed predisposition to KCS worldwide, most national eye panels for breeding dogs do not include mandatory quantitative tear testing within their standardised eye examination in any breed. The authors of the present study suggest that inclusion of quantitative tear testing as part of the standardised eye examination of breeding dogs in those breeds with known predispositions to KCS would be sensible to assist with reduction in the incidence of KCS in these breeds. The six breeds suggested by the authors of the present study are ideal candidates to begin this process. It would also appear sensible to recommend consideration of quantitative tear testing in these predisposed breeds within their individual Breed Health And Conservation Strategy Plans that are devised by the UK Kennel Club in conjunction with the relevant breed clubs (The Kennel Club 2020a).
Dogs with acute KCS often present clinically with deep corneal ulceration amid few other clinical signs, prompting a previous recommendation to perform quantitative tear testing in all animals with a history of conjunctivitis, even when patients are young and their clinical signs are mild (Sanchez et al. 2007). The findings of the present study show that advancing age was strongly associated with increasing adjusted odds of KCS. This is not surprising if one takes into account that the disease will worsen over time while it remains undiagnosed ). Moreover, mean tear readings measured with Schirmer's tear test −1 naturally decrease with age by 0.4 mm/minute for every year  British Small Animal Veterinary Association increase in age (Hartley et al. 2006). Based on these findings, the current authors suggest that it would be sensible to extend the recommendation of Sanchez et al. (2007) beyond just conjunctivitis to now also recommend that the ocular component within yearly preventive health examinations should include a quantitative tear test in those breeds predisposed to KCS, even for animals without other suggestive clinical signs of KCS. It has been suggested previously that dogs that present with KCS from breeds with prominent eyes as part of their conformation often also show keratitis in addition to conjunctivitis . Breeds with prominent eyes, such as shih-tzu and cavalier king Charles spaniel, were reported to have a significantly higher risk of corneal ulcerative disease associated with KCS compared to the other breeds in one study (Sanchez et al. 2007). Moreover, a recent large study of corneal ulcerative disease in dogs under primary veterinary care in the UK reported that pugs and boxers as well as brachycephalic dogs and spaniels were predisposed to corneal ulceration (O'Neill et al. 2017). That paper suggested that breeding reforms that focused on reducing the degree of exaggerated periocular conformation in those breeds might help reduce the incidence of corneal ulcerative disease. The present study reports that small breeds, as well as purebreds, brachycephalic dogs and spaniels, each have a higher risk of KCS. These breed and breed-related risk factors closely overlap with those for corneal ulcerative disease (O'Neill et al. 2017, Sanchez et al. 2007. Taken together, these findings are strongly suggestive that facial conformations with excessively large interpalpebral aperture, ectropion of the lower eyelid and/ or very shallow orbits predispose to corneal ulcerative disease. Therefore, it appears sensible to consider options to breed away from extremes of facial conformation when developing breeding strategies to reduce ocular problems in breeds predisposed to KCS (The Kennel Club 2020a). The present study identified that dogs weighing at or above the mean for that breed/sex were associated with 1.25 times increased adjusted odds of KCS. It is worth noting that dogs with a bodyweight at or above the mean for their breed/sex are not necessarily overweight/obese, although this may be a contributory factor in some instances. Excess bodyweight has been associated with an increased risk for inherited conditions such as cruciate ligament rupture, hip dysplasia, and hypothyroidism (German 2006, Lund et al. 2006, Simpson et al. 2019) as well as for certain types of cancer (Weeth et al. 2007). Although neutering has often been considered to be associated with subsequent weight gain in dogs, there are still some conflicting results regarding this effect (Reichler 2009). A retrospective study reported increased expression of some heritable conditions in neutered animals but further reported that this effect was not consistent across all heritable conditions (Belanger et al. 2017). Another study reported that the association between neutering and the expression of an inherited condition was a direct effect of the neutering and was not due to other factors, such as an alteration of weight metabolism (Oberbauer et al. 2019). Although the present study reported an association between bodyweight and increased adjusted odds for KCS, it did not identify any association between neuter status and KCS, or between sex and KCS. A previous study has proposed that a predisposition to KCS in female neutered dogs could be due to the loss of protective hormones (Kaswan et al. 1991). However, some studies have failed to identify evidence of a sex predisposition for KCS in dogs (Aguirre et al. 1971, Kaswan et al. 1985 while another study reported either male or female predisposition to KCS depending on the breed, and that entire (unneutered) animals were more common among cases (Sanchez et al. 2007). Given the conflicting results between studies, predispositions to KCS related to sex, neuter status or bodyweight remain difficult to quantify and explain. However, based on the findings of the present study, a sensible approach would be to extend the clinical rationale for the general primary-care recommendation for good bodyweight control by informing owners of a potentially higher predisposition to KCS in small breeds with excessive weight gain (Ward et al. 2018).
The application of primary-care veterinary data for epidemiological research has several limitations that have been previously reported (O'Neill et al. 2014a(O'Neill et al. , 2019. These include issues related to the re-use of data that were not recorded for research purposes, reliance on accurate note-taking, differing clinical beliefs between clinicians and frequent high levels of missing data. In the specific context of the current study, an additional limitation is that primary care veterinary surgeons may have different levels of experience in the general medicine of ophthalmic practice. Consequently, less experienced veterinarians may be less alert to the clinical presentation of KCS or, conversely, might overinterpret the findings of their ophthalmic examination; either way, this may result in misclassification of KCS cases. Another limitation is that the searching methods used to identify all cases in the current large database depended on the use of key words that could potentially have under-detected true cases. This would have the effect of under-reporting the prevalence and incidence but should